Tri-domain Bifunctional Inhibitor of Metallocarboxypeptidases A and Serine Proteases Isolated from Marine Annelid Sabellastarte magnifica*

Background: Several protein inhibitors of metallocarboxypeptidases have already been described. Results: We have characterized of a tri-domain inhibitor from Sabellastarte magnifica, recombinant forms and truncated variants. Conclusion: The whole tri-domain is required for full inhibition of metallocarboxypeptidases A. Monodomains are designed to inhibit serine proteases. Significance: The first reported multidomain protein inhibitor of metallocarboxypeptidases is also able to act on another mechanistic class of proteases (serine-type). This study describes a novel bifunctional metallocarboxypeptidase and serine protease inhibitor (SmCI) isolated from the tentacle crown of the annelid Sabellastarte magnifica. SmCI is a 165-residue glycoprotein with a molecular mass of 19.69 kDa (mass spectrometry) and 18 cysteine residues forming nine disulfide bonds. Its cDNA was cloned and sequenced by RT-PCR and nested PCR using degenerated oligonucleotides. Employing this information along with data derived from automatic Edman degradation of peptide fragments, the SmCI sequence was fully characterized, indicating the presence of three bovine pancreatic trypsin inhibitor/Kunitz domains and its high homology with other Kunitz serine protease inhibitors. Enzyme kinetics and structural analyses revealed SmCI to be an inhibitor of human and bovine pancreatic metallocarboxypeptidases of the A-type (but not B-type), with nanomolar Ki values. SmCI is also capable of inhibiting bovine pancreatic trypsin, chymotrypsin, and porcine pancreatic elastase in varying measures. When the inhibitor and its nonglycosylated form (SmCI N23A mutant) were overproduced recombinantly in a Pichia pastoris system, they displayed the dual inhibitory properties of the natural form. Similarly, two bi-domain forms of the inhibitor (recombinant rSmCI D1-D2 and rSmCI D2-D3) as well as its C-terminal domain (rSmCI-D3) were also overproduced. Of these fragments, only the rSmCI D1-D2 bi-domain retained inhibition of metallocarboxypeptidase A but only partially, indicating that the whole tri-domain structure is required for such capability in full. SmCI is the first proteinaceous inhibitor of metallocarboxypeptidases able to act as well on another mechanistic class of proteases (serine-type) and is the first of this kind identified in nature.

This study describes a novel bifunctional metallocarboxypeptidase and serine protease inhibitor (SmCI) isolated from the tentacle crown of the annelid Sabellastarte magnifica. SmCI is a 165-residue glycoprotein with a molecular mass of 19.69 kDa (mass spectrometry) and 18 cysteine residues forming nine disulfide bonds. Its cDNA was cloned and sequenced by RT-PCR and nested PCR using degenerated oligonucleotides. Employing this information along with data derived from automatic Edman degradation of peptide fragments, the SmCI sequence was fully characterized, indicating the presence of three bovine pancreatic trypsin inhibitor/Kunitz domains and its high homology with other Kunitz serine protease inhibitors. Enzyme kinetics and structural analyses revealed SmCI to be an inhibitor of human and bovine pancreatic metallocarboxypeptidases of the A-type (but not B-type), with nanomolar K i values. SmCI is also capable of inhibiting bovine pancreatic trypsin, chymotrypsin, and porcine pancreatic elastase in varying measures. When the inhibitor and its nonglycosylated form (SmCI N23A mutant) were overproduced recombinantly in a Pichia pastoris system, they displayed the dual inhibitory properties of the natural form. Similarly, two bi-domain forms of the inhibitor (recombinant rSmCI D1-D2 and rSmCI D2-D3) as well as its C-terminal domain (rSmCI-D3) were also overproduced. Of these fragments, only the rSmCI D1-D2 bi-domain retained inhibition of metallocarboxypeptidase A but only partially, indicating that the whole tri-domain structure is required for such capability in full. SmCI is the first proteinaceous inhibitor of metallocarboxypeptidases able to act as well on another mechanistic class of proteases (serine-type) and is the first of this kind identified in nature.
Metallocarboxypeptidases (CPs) 5 are an important class of enzymes that catalyze the hydrolysis of peptide bonds at the C terminus of peptides and proteins. Besides a role in digestive protein degradation, these enzymes are also key elements of selective proteolysis-regulated physiological processes such as blood coagulation/fibrinolysis, inflammation, prohormone and neuropeptide processing, local anaphylaxis, and insect/plantattack/defense strategies, among others (1,2). The biological actions of many proteases are controlled by their interaction with specific proteinaceous inhibitors. Unlike endoproteases, for which numerous examples of protein inhibitors have been reported, naturally occurring metallocarboxypeptidase inhibitors are somewhat limited, and so far, they have only been identified in Solanacea, tomato, and potato (PCI) (3)(4)(5), the intestinal parasite Ascaris suum (ACI, Ascaris carboxypeptidase inhibitor) (6,7), the medicinal leech Hirudo medicinalis (LCI) (8), the ticks Rhipicephalus bursa (TCI) (9) and Hemaphysalis longicornis (H1TCI) (10), and in rat and human tissues (latexin or endogenous carboxypeptidase inhibitor (ECI)) (11,12).
The mechanisms of the inhibitory actions of PCI, LCI, and TCI on CPs rely upon interaction of their C-terminal tail with the active site cleft of the enzyme in a manner that mimics substrate binding (2,5,8,13). Additionally, TCI anchors to the surface of CPs of the A/B-type in a double-headed manner not observed for the other protein inhibitors (13). However, the C-tail of mammalian tissue protein inhibitors does not seem to be a suitable substrate for CPs; such proteins interact with the enzymes through one loop located at the interface of their two subdomains (14). In addition, the pro-regions of procarboxypeptidases, which fold as independent globular domains, position their internal inhibitory loop on the active site cleft of the enzyme rendering the enzyme inhibited (15,16). All these inhibitors are specific for the A/B metallo-CP subfamily, regardless of their substrate preferences (2).
In general, protein inhibitors of proteases belonging to different mechanistic classes are uncommon. Such inhibitors may feature one or more inhibitory domains, such as SHPI-1 (with one BPTI/Kunitz-type domain), which is able to inhibit serine, cysteine, and aspartic proteases (17), or equistatin (with three thyroglobulin-1 domains), which inhibits cysteine and aspartic proteases (18,19), among others. However, there have been no descriptions to date either of a multifunctional inhibitor able to inhibit CPs and proteases belonging to distinct mechanistic classes nor of inhibitors of CPs with a typical Kunitz structure with the capacity to inhibit several serine proteases.
Among the available natural sources of protease inhibitors, one of the most attractive and rather unexplored is the marine fauna, especially invertebrates (including numerous phyla, genera, and species). Several such inhibitors capable of independently inhibiting proteases of different mechanistic classes have been described, particularly in the phyla Cnidaria (17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27), Mollusca (28 -31), and Annelida (32,33). However, no inhibitors of CPs of this type had been described in such sources until the present. In a preliminary survey, we detected the remarkable CP inhibitory capacity of tentacle crown preparations of the marine annelid Sabellastarte magnifica, also known as the "magnificent feather duster" (Phylum Annelida and class Polychaeta) (34).
Here, we describe the isolation, characterization, cDNA cloning, and sequence analysis of a novel tight-binding metallo-CP inhibitor found in S. magnifica. This inhibitor is bifunctional, also acting on serine proteases, featuring three BPTI/ Kunitz domains not present in other proteinaceous CP inhibitors. When SmCI and its nonglycosylated form were overexpressed in a Pichia pastoris system, the recombinant forms showed similar bifunctional properties to the natural form. In addition, expression and characterization of bi-domains and third domain recombinant inhibitors allowed us to go deeply into the kinetic behavior of SmCI against the enzymes that it can inhibit. Our findings expand our existing knowledge and repertoire of inhibitors of both CP and Kunitz-type serine protease. Our findings also provide insight into the distinctive features of such molecules in the still quite unexplored world of marine invertebrates as a potential rich and diverse source of new substances of biotechnological interest.
Purification of SmCI-The marine invertebrate was collected at north of Havana, Cuba, and was taxonomically identified by specialists of the Cuban National Institute of Oceanology. The tentacle, or feathered, crowns of the animals were separated from the body, homogenized (2:1 v/w), and centrifuged. The supernatant was clarified by heating at 60°C for 20 min and centrifuged. The heated extract was loaded in three steps onto a CPA-glyoxyl-agarose column (0.9 ϫ 5.5 cm) prepared as 1.3 mg of immobilized CPA per ml of gel according to the general procedure described for other enzymes (35), with some modifications. Unbound proteins were eliminated by washing the column with a sufficient quantity of equilibration buffer (0.05 M Tris-HCl, 0.5 M NaCl, 10 Ϫ5 M ZnCl 2 , pH 7.0). Proteins with CPA inhibitory activity were eluted by increasing the pH to 10.4 through the addition of 0.05 M glycine, 0.04 M NaOH buffer at a linear flow rate of 24 cm/h. Inhibitor-containing fractions were lyophilized and subjected to reverse phase HPLC (Vydac C8 column). The homogeneity of the purified inhibitor was verified by SDS-PAGE (36) and MALDI-TOF mass spectrometry. The SDS-polyacrylamide gel was stained with Coomassie Blue R-250. Prestained molecular weight standards were used.
Active Concentrations of Trypsin and Inhibitor-The active pancreatic trypsin concentration was determined by titration with the standard solution of p-nitrophenyl-p-guanidinobenzoate (4.9 ϫ 10 Ϫ4 M in the assay) (44). The active inhibitor concentrations were determined by titrating with increasing concentrations of trypsin (from 0.19 ϫ 10 Ϫ7 to 1.7 ϫ 10 Ϫ7 M in the assays) using a constant concentration of the inhibitor (8.0 ϫ 10 Ϫ8 M) and a preincubation time of 10 min, under conditions of E o /K i ϭ 100. Residual activity was assayed using 1 mM BAPA, and the active inhibitor concentration was determined at the equivalence point ([E t ] ϭ [I t ]) indicated in a plot of residual activity against enzyme concentration.
Dissociation Constants (K i )-Inhibition constants (K i ) for the complexes formed by the inhibitors with the different enzymes were determined using a described method for tight binding inhibition (45). Different amounts of inhibitor were preincubated for 10 min with the enzymes under conditions of E o /K i ϭ 10. At each concentration, residual activity (v i ) was measured against the specific substrate (using a substrate concentration equivalent to 1 K m for each enzyme). The experimental points were adjusted to the equation described for tight binding mechanisms (46) by nonlinear fitting using the GraphPad Prism 5.0 package. True K i values were calculated using the equation: , incorporating the [S 0 ] used and the K m value for each enzyme.
Protein Concentration-Protein concentrations were determined by the bicinchoninic acid method (47) using the BCA kit (Pierce) and bovine serum albumin as standard. Mixed protein concentrations were determined by measuring absorbance at 280 nm assuming A 280 nm (1%) ϭ 10. Molecular Mass Determination-The molecular mass of peptides and proteins was established by MALDI-TOF mass spectrometry (MS) using an Ultraflex TOF-TOF instrument (Bruker, Bremen, Germany). Ionization was achieved with a 337-nm pulsed nitrogen laser, and spectra were acquired in the linear positive ion mode applying a 19-kV acceleration voltage. Sinapinic acid (3,5-dimethoxy-4-hydroxycinnamic acid) and ␣-cyano-4-hydroxycinnamic acid were used as matrices for the analysis of proteins or peptide fragments. Samples were prepared by mixing equal volumes of a saturated solution of the matrix in aqueous 30% acetonitrile with 0.1% trifluoroacetic acid (v/v). A 1-l aliquot of this mixture was then spotted on the sample slide and allowed to evaporate to dryness.
Reduction, Carboxymethylation, and Proteolytic Cleavage-An aliquot of 200 g of SmCI was denatured in 6 M guanidine hydrochloride and then reduced and carboxymethylated with dithiothreitol and iodoacetic acid, respectively (9). The solution was diluted three times with milli Q-grade water and treated with trypsin, endoprotease LysC, and V8 endoprotease (GluC endoprotease from Staphylococcus aureus) for 24 h at 37°C at a 10:1 ratio (w/w). For endoprotease AspN (from a Pseudomonas fungi mutant) digestion, the inhibitor was diluted six times and incubated with the enzyme at a 30:1 ratio (w/w) for 24 h at 37°C. Reactions were stopped by adding the same volume of 0.1% TFA in water. Peptide fragments were purified using a Vydac C18 column in the HPLC system. The carboxymethylated inhibitor and isolated peptide fragments were analyzed by automatic amino acid sequencing on an Edman-based Beckman LF3000 protein sequencer and by MALDI-TOF MS.
Detection of cDNAs Encoding SmCI-Total RNA was isolated from tentacle crown homogenates using the Nucleospin kit, and poly(A) ϩ RNA was purified using the Nucleotrap kit both according to the manufacturer's instructions. The first strand of SmCI cDNA was synthesized using the adaptor oligonucleotide R 0 R 1 -dT (R 0 , 5Ј-CCGGAATTCACTGCAG-3Ј; R 1 5Ј-GT-ACCCAATACGACTCACTATAGGGC-3Ј) and avian myeloblastosis virus reverse transcriptase according to the supplier's protocols. For cloning the SmCI cDNA, two degenerated oligonucleotides were designed based on its N-terminal sequence, S1(8 -17), 5Ј-GCNGAYTGYGGNCARTGYCANGCNTAYA-T-3Ј (residues 8 -17 from the N-terminal sequence), and S2(13-22) (amino acid residues 13-22) 5Ј-TGYCANGCN- For the first PCR round, SmCI cDNA was amplified using oligonucleotide S1 (8 -17) and the adaptor oligonucleotide R 0 . PCR was conducted as 40 cycles each at 94°C for 1 min, annealing at 48°C for 1 min, and extension at 72°C for 2 min. PCR mixtures were 20-fold diluted and re-amplified using the S2(13-22) and R 1 nested oligonucleotides. PCR products were separated by electrophoresis on 2% agarose gels. The main nested PCR product was recovered from the gel and cloned into pGEM-T-easy vector to generate the pGEM-SmCI S1/S2 construct (corresponding to SmCI amino acid residues from 13 to 165).
Nucleotide Sequencing and Computational Analysis of Sequence Data-The pGEM-SmCI S1/S2 clones generated were sequenced using the SP6 and T7 promoter forward and reverse primers in a DNA sequencer (PerkinElmer Life Sciences). Alignments of nucleotides and amino acid residues were carried out using VECTOR NTI as implemented by the VECTOR NTI Suite 9 program of the InforMax 2003 package. PSI-BLAST and PFAM (48,49) were used to search nonredundant databases. Multiple sequence alignment was performed by combining Clustal X (50), the protein sequences identified by PSI-BLAST, and the disulfide bridge information derived from the resolved crystal structures using DSSP (51).
Deglycosylation Assay-Samples were deglycosylated with N-glycosidase F. SmCI prepared at 1 mg/ml in 5 mM Tris-HCl buffer, pH 8.0, was incubated with an appropriate volume of N-glycosidase F (1 unit/l) to achieve a final ratio of 100:1 v/v. The reaction was left to run for 24 h at 37°C. The deglycosylation process was monitored by MALDI-TOF MS.
Heterologous Expression of the Recombinant Inhibitors-rSmCI, rSmCI N23A (a nonglycosylated form), the two bi-domains rSmCI D1-D2 and rSmCI D2-D3, and its third domain (rSmCI-D3) were overexpressed in the P. pastoris system. First, the SmCI encoding sequence was completed by PCR using two specific sense gene primers (S3(6 -18) and S4(1-9)). These primers were designed from the N-terminal sequence of the protein, S3(6 -18) (residues 6 -18) 5Ј-TTGCCAGCTGATAGAGGTCAATGT-ACGGCCTACATTCCC-3Ј and S4(1-9) (residues 1-9) 5ЈTCTCTCGAGAAAAGAATTTCTGTTTGTGATTTGCC-AGCTGAT-3Ј. As the antisense primer we used S5(163-165), which corresponds to the SmCI C-terminal sequence (residues 163-165) 5Ј-CCTTCGCGGCCGCCTAGCAAGCATT-3Ј. These primers included the restriction sites for XhoI and NotI (underlined) to clone the gene into the pPICZ␣A vector, and the Kex 2 cleavage site (sense primer) and stop codon (antisense primer) are shown in boldface type. PCR was conducted under the same conditions as described above using pGEM-SmCI S1/S2 vector as template. SmCl gen was removed from pGEM-T easy vector using XhoI and NotI restriction sites and cloned into pPICZ␣A vector. The new generated vector, designated pPICZ␣A-SmCI, was transformed in the KM71H strain (Mut s phenotype) of P. pastoris following the manufacturer's protocol (Invitrogen). Several colonies on the YPDS agar/Zeocin plate were screened for small scale expression (following the instructions provided in the Invitrogen catalogue). The most promising clone was then grown in BMGY medium at 28°C for 1 day, and induced in BMMY medium for 48 h following the procedure provided by the manufacturer (Invitrogen).
Purification and Partial Characterization of the Recombinant Inhibitor Forms-rSmCI, rSmCI N23A, rSmCI D1-D2, and rSmCI D2-D3 were purified using similar protocols. The supernatants obtained after fermentation were loaded onto an Streamline Direct HST column (2.0 ϫ 25.0 cm) equilibrated with 100 mM sodium citrate, pH 4.0 (buffer A1), using an ÄKTAprime system (GE Healthcare). Unbound proteins were eliminated by washing the column with 20 mM sodium citrate, pH 4.0 (buffer A2). The recombinant protein was eluted using a gradient of 25-100% buffer B (150 mM Tris-HCl, pH 8.0) in 20 column volumes at 10 ml/min. Fractions showing CPA and/or trypsin inhibitory activity were applied to an anion exchange column of HiTrapTM Q-Sepharose FF (1 ml) (General Electric), previously equilibrated with 20 mM Tris-HCl, pH 8.0, buffer at a flow rate of 1 ml/min. After washing with the equilibration buffer (6 column volumes), bound inhibitor was eluted with a linear gradient from 0 to 45% of 0.5 M sodium chloride in 20 mM Tris-HCl, pH 8.5, in 20 column volumes at the same flow rate. rSmCI-D3 was purified by combining the reverse phase in the Sep-Pak C18 cartridge with the anion exchange step in Q-Sepharose.
The molecular weights, N-terminal sequences, and inhibitory activity against CPA, CPB, trypsin, and pancreatic elastase of the recombinant forms were assessed as described above for the natural protein. K i values of the inhibitors against different enzymes were determined using the same kinetic strategy and substrates employed for the natural inhibitor. For the pancreatic elastase inhibitory assay, N-Suc-(Ala) 3 -pNA was used as substrate in case of rSmCI (41).

RESULTS
Isolation and Purification of SmCI-The proteinaceous inhibitor was isolated from the tentacle crown of S. magnifica by both affinity and reverse phase chromatography. The inhibitory activity recovered after heat treatment at 60°C (20 min) was 92% indicating the stability of SmCI at this temperature. The heated extract was loaded in three steps onto an affinity column (CPA-glyoxyl-Sepharose). This procedure allows the application of a larger amount of inhibitor than standard methods (one application) (Fig. 1A). The inhibitor was strongly bound to the immobilized enzyme and eluted by increasing the pH of the buffer to 10.4. The yield, based on the inhibitory capability on CPA, was 113%, and purification was 102-fold with respect to the crude extract. The main component of the affinity fraction applied to a reversed phase HPLC was purified to homogeneity in appropriate conditions for its molecular and kinetic characterization. SDS-PAGE of the purified material showed a single protein band around 20 kDa (Fig. 1B). This result is consistent with a molecular mass of 19.69 kDa determined by MALDI-TOF MS (Fig. 1C).
Inhibitory Activity and Selectivity of SmCI-Purified SmCI was able to inhibit proteases pertaining to two different mechanistic classes, metallocarboxypeptidases (A subtype, but not B) and serine proteases (trypsin, pancreatic elastase and chymotrypsin). In contrast, using up to a 10-fold molar excess of inhibitor, no blocking action was detected on the activities of papain or pepsin, two standard models for cysteine and aspartic protease.
Equilibrium dissociation constants (K i ) for the interaction of SmCI with CPs, trypsin, elastase, and chymotrypsin were examined by enzyme kinetics analysis. Preincubation of the inhibitor with serine proteases over varying periods of time did not affect its inhibitory capacity, suggesting that SmCI quickly binds to these enzymes, although CP inhibitory activity needed at least a minute for equilibrium to be reached before its action. In all cases, the concentrations of both protease and inhibitor were sufficiently low to generate concave inhibition curves (data not shown) indicative of reversibility (45,46).
Values of K i , adjusted according to the substrate-induced dissociation recorded for different amounts of each substrate (results not shown), indicated a strong inhibitory action on bovine pancreatic CPA, trypsin, and pancreatic elastase, with K i values between the 10 Ϫ8 and 10 Ϫ9 M range (Table 1). In contrast, no inhibition capacity for CPB was observed, and chymotrypsin was inhibited in smaller measure, with a K i value in the micromolar range (1.83 Ϯ 0.92 M). Similar behavior was observed toward CPA and CPB from other species, including human forms of the enzymes (data not shown).
Primary Structure of SmCI-Edman degradation of purified natural SmCI indicated the following N-terminal sequence for the inhibitor: NH 2 -ISVCDLPADRGQCTAYIPQWFF (22 residues released). MALDI-TOF MS revealed the incorporation of 18 carboxymethyl groups when the inhibitor was reduced and carboxymethylated, although these were absent when the intact molecule was treated, suggesting that SmCI contains 18 cysteine residues involved in disulfide bonds. The reduced and carboxymethylated inhibitor was fragmented with trypsin, Lys-C, Asp-N, and Glu-C endoproteases. The resulting overlapping peptides (Fig. 2) analyzed by Edman degradation covered almost the entire sequence of SmCI. The whole sequence was subsequently completed and confirmed by cDNA cloning (see below).
Cloning and Sequence Analysis of SmCI-Using RT-PCR and nested PCR, we obtained a partial SmCI cDNA sequence from RNA that had been previously isolated as a template for the RT reaction. The product of the first PCR (S1(8 -17) and R 0 primers) appeared on agarose gel as a single band of around 700 bp. Nested PCR rendered a prominent PCR product of about 600 bp, which was then purified and subcloned to generate a pGEM-SmCI S1/S2 construct. Ten clones were checked by restriction analysis and sequenced using the SP6 and T7 promoter forward and reverse primers. Sequence analysis of the plasmids showed the presence of 495 bp corresponding to the SmCI cDNA sequence. By translating this nucleotide sequence, we were able to complete and confirm the SmCI protein sequence from residue 13 to the C terminus of the protein (Fig.  3).
The whole SmCI protein sequence derived from the nucleotide sequence includes 165 amino acids. The protein is rich in glycine (16 residues), glutamic acid (17 residues), and cysteine (18 residues); the latter participates in nine disulfide bridges. This sequence showed a theoretical molecular mass of 18.649 kDa, revealing a difference of 1054 Da with respect to the experimental feature determined by MALDI-TOF MS. Considering the presence of a potential N-glycosylation site (N/X/T) at Asn 23 , deglycosylation assays were followed by MALDI-TOF MS. A difference of 1054 Da was found between the intact and deglycosylated form of the inhibitor, indicating that SmCI is an N-glycosylated protein.
Computer analysis of the SmCI sequence revealed a high homology with BPTI/Kunitz proteins. This facilitated a preliminary prediction of the disulfide bond pattern, fully conserved in the members of this family (52). PFAM analysis (49) indicated the presence of three BPTI/Kunitz domains in the SmCI sequences with PFAM E values between e Ϫ27 and e Ϫ22 for each domain. NCBI nonredundant sequence database searches using PSI-BLAST revealed that SmCI most resembled proteins with more that one Kunitz domain in their structure such as follows: the inter-␣-trypsin inhibitor of different species (horse, goat, and sheep) with E values ϭ 3.3 ϫ e Ϫ19 ; human, bovine, pig, mouse and rat AMBP protein precursor (inter-␣-trypsin inhibitor light chain, bikunin) with E values between e Ϫ18 and e Ϫ17 ; human, mouse, and rat tissue factor pathway inhibitor 1 precursor and tissue factor pathway inhibitor 2 precursor (TFPI1 and TFPI2, respectively), which have three Kunitz domains, with E values of e Ϫ15 and e Ϫ10 , respectively. Furthermore, similarities (ϳe Ϫ7 to e Ϫ12 ) were also found with several domains of Kunitz proteins and with Kunitz-like serine protease inhibitors of species showing a different phylogenetic distribution, e.g. the sea anemone inhibitors ShPI-1 and ShPI-2 (e Ϫ10 ) from Stichodactyla helianthus; Kunitz-type protease inhibitors AXPI-1 and AXP2 (e Ϫ10 and e Ϫ7 , respectively) from Anthopleura xantho- grammica; Kunitz-type protease inhibitor 5 from Anemonia sulcata (e Ϫ10 ); and snake venom basic protease inhibitors (with E values between e Ϫ7 and e Ϫ11 ), among others.
A diagram of the SmCI inhibitor's primary structure and disulfide bridges, according to the BPTI/Kunitz disulfide pattern (Cys I -Cys VI , Cys II -Cys IV , and Cys III -Cys V ) (Prosite PS00280) (52), is provided in Fig. 4. The three domains display the basic structure and consensus sequence of the members of this family based on homology considerations. Potential P3-P3Ј enzyme recognition sites around the P1-P1Ј reactive site are depicted for the three domains in this figure.
Modeling the Three-dimensional Structure of SmCI Domains-Taking into account the high degree of similarity between SmCI and BPTI/Kunitz protein sequences along with the absence of three-dimensional structures established for proteins with three Kunitz domains in their structure, we comparatively modeled the individual SmCI domains. The threedimensional crystal structures used as templates were selected on the basis of their three-dimensional JURY scores (above 45%) and high resolution (below 2.5 Å) (53). To construct specific models for SmCI-D1, SmCI-D2, and SmCI-D3, we used as template the structure of the second domain of human TFPI1 (Protein Data Bank code 1TFX), the second domain of human bikunin (Protein Data Bank code 1BIK), and the first domain of human TFPI2 (Protein Data Bank code 1ZRO), respectively.
Based on these templates, we generated 100 models for each domain using the MODELLER program (54) and quality packing values below Ϫ5.0. Fig. 5 shows the best calculated models for each domain with average values for the quadratic medium deviation of 100 models equal to 0.223 Ϯ 0.04, 0.199 Ϯ 0.031, and 0.252 Ϯ 0.024 for SmCI-D1, SmCI-D2, and SmCI-D3, respectively. Each domain is folded in a central anti-parallel ␤-sheet with one ␣-helix toward the C terminus, which is typical of BPTI/Kunitz proteins. However, according to the models, the domains differ in their N terminus, which lack a regular structure in the first two domains and feature an ␣-helix in the third domain, instead of the 3 10 helix commonly found in this family of proteins.
To clarify the potential mechanism of SmCI's inhibitory effect on CPs, we compared the C-terminal sequence of its third domain to those of CP protein inhibitors in which these regions are involved in their inhibitory mechanisms (Fig. 6). SmCI showed no similarities in this region with such protein inhibitors. It should be highlighted that the C-tail of SmCI shows limited accessibility and flexibility because of the presence of a cysteine residue at the end, involved in a disulfide bridge in this third Kunitz domain. These characteristics potentially hinder its binding to CPs. This rationale was confirmed when we found that the third domain and the bi-domain D2-D3 expressed as a recombinant forms were unable to inhibit CPA (see Table 1). With regard to inhibitory actions on serine proteases, alignments of reactive site sequences of the individual domains with those of several typical BPTI/Kunitz inhibitors (Table 2) revealed several differences between the three domains, which could be related to its individual inhibitory capacities against different serine proteases. The third domain bears the typical reactive site for trypsin inhibition with a lysine residue at P1 and     a glycine at P1Ј, consistent with the inhibitory activity toward trypsin observed for this domain.

TABLE 1 K i values (mol/liter) of the tri-domain forms of SmCI and all its recombinant domains against trypsin, pancreatic elastase, and pancreatic (p) CPA1
Heterologous Expression and Characterization of Recombinant Inhibitors-Using the cloning strategic described under "Experimental Procedures," we produced the different recombinant forms of the inhibitor. By combined rSmCI overexpression and purification, we produced a fraction that was highly enriched in the active recombinant protein showing several signals around 20 kDa in MALDI-TOF MS spectra (results not shown). These signals probably corresponded to different gly-cosylation grades of the protein, frequently described in the P. pastoris system (65). Automatic Edman degradation analysis revealed a single N-terminal sequence (ISVCDLPADR) matching the natural SmCI N-terminal sequence.
In contrast, rSmCI N23A, made to avoid the glycosylation, showed a single peak with a molecular mass of 18,606 Da by MALDI-TOF MS, which correspond to the theoretical molecular mass for this alanine mutant determined by ExPASy server. The asparagine residue mutated was confirmed by Edman sequencing.
Purified rSmCI and rSmCI N23A revealed the same inhibitory capacity and specificity toward proteases as the natural inhibitor. Dose-effect assays of the inhibitors against metallocarboxypeptidase A showed a concave kinetic behavior curve for this enzyme (Fig. 7A). The recalculated K i values (considering substrate dissociation) against CPA, trypsin, and pancreatic elastase are summarized in Table 1. The recombinant tri-domains (glycosylated and nonglycosylated) and natural forms showed similar K i values against pancreatic CPA and serine proteases confirming their bifunctional properties similar to the natural SmCI inhibitor.
When the bi-domain proteins rSmCI D1-D2 and rSmCI D2-D3 were overexpressed and purified, we got two proteins with molecular masses of 11,949 and 12,549 Da, respectively, and N-terminal sequences that correspond with both proteins. rSmCI D1-D2 was able to inhibit trypsin and pancreatic elastase with K i values of 1.20 ϫ 10 Ϫ7 and 5.55 ϫ 10 Ϫ8 M, respectively; and for CPA, rSmCI D1-D2 showed K i ϭ 1.96 ϫ 10 Ϫ6 M, indicating less strength of inhibition against the three enzymes than that obtained for tri-domain inhibitor forms ( Fig. 7B and Table 1). rSmCI D2-D3 and rSmCI-D3 only inhibited trypsin with K i values of 1.4 ϫ 10 Ϫ9 and 3.05 ϫ 10 Ϫ8 M, respectively, displaying null inhibitory activity on other serine proteases or on CPs (Table 1).

DISCUSSION
In this study, we identified and characterized (at both the protein and DNA levels) a novel metallocarboxypeptidase inhibitor (SmCI) isolated from the tentacle crown of the marine annelid S. magnifica. By using several complementary strategies, we were able to establish its complete SmCI protein sequence. In addition, an N-glycosylation site at residue Asn 23 was predicted and confirmed experimentally in the natural form by deglycosylation assays.
Functionally, SmCI is a tight-binding inhibitor that inhibits both carboxypeptidase A and serine proteases, such as trypsin and pancreatic elastase, with equilibrium dissociation constants in the nanomolar range, whereas chymotrypsin is inhibited with a K i value only in the micromolar range. These functional characteristics were reproduced by the recombinant tri-domain forms (glycosylated and nonglycosylated), confirming that they are correctly folded and functionally equivalent to the protein isolated from the tentacle crown. Nevertheless, the results obtained in this work suggest that in case of SmCI N23A, a nonglycosylated form of the protein by elimination of the sugar moiety, it displays an increment of its trypsin inhibitory capability, probably reflecting a better enzyme-inhibitor fit. Aligned inhibitors, SmCI (PCPI_SABMA), carboxypeptidase inhibitor from S. magnifica; TCI (TCI1_RHIBU), carboxypeptidase inhibitor from R. bursa (9); HITCI (A8C364_HAELO), carboxypeptidase inhibitor from H. longicornis (10); MPCI (MCPI_SOLLC) carboxypeptidase inhibitor from Solanum lycopersicum (6); PCI (MCPI_SOLTU) carboxypeptidase inhibitor from Solanun tuberosum (3,4); LCI (MCPI_HIRME) carboxypeptidase inhibitor from H. medicinalis (8); Ascaris carboxypeptidase inhibitor (ACI) (ICAA_ASCSU), carboxypeptidase inhibitor from A. suum (7). Similar or identical residues among the C-terminal amino acid residues of the inhibitors are shaded.

TABLE 2
Amino acid sequences around the reactive site residues of selected BPTI/Kunitz inhibitors Aligned inhibitors are as follows: SmCI-D1, SmCI-D2, and SmCI-D3, first, second, and third domains of the inhibitor from S. magnifica; AXP1_ANTAF and AXP2_ANTAF from inhibitors 1 and 2 from the sea anemone A. xanthogrammica (27); IP52_ANESU from main inhibitor from the sea anemone A. sulcata (24); ISH1_STOHE and ISH2_STOHE, inhibitor 1 and 2 from the sea anemone S. helianthus (17,55); AceKI-1, inhibitor from the nematode Ancylostoma ceylanicum (56); Q8MtR6_HAEIR, inhibitor from the cattle tick Haematobia irritans (57)  Consistent with its serine inhibitory capacity, SmCI shows remarkable homology with proteins of the BPTI/Kunitz family, whose members include inhibitors of serine proteases but not of metallocarboxypeptidases. By aligning the whole SmCI sequence with other Kunitz domains, we observed that SmCI includes three such domains, each of which features the basic structure of the members of this family, including the same spacing of the six conserved cysteine residues, as well as the consensus sequence FXYXGCXGNXN around the fourth cysteine residue (52). The three-dimensional structures generated for each domain by comparative modeling also displayed the typical structural features of BPTI/Kunitz inhibitors (66 -67), as the most representative canonical protease inhibitors.
The BPTI/Kunitz protease inhibitor family, exemplified by bovine pancreatic trypsin inhibitor, has been the subject of intense investigations. Numerous variants of these protein inhibitors have been isolated from different tissues and animal sources, spanning from turtle egg white and snake venom to the major organs of ruminant mammals (68). The specificity of each of these protease inhibitors is mainly dependent on the nature of the residue at position P1 of the active site (69). Effectively, two of SmCI domains (the second and third) have basic P1 residues, such as arginine and lysine, respectively, as do most BPTI/Kunitz trypsin inhibitors (70); although the first SmCI domain bears a threonine, a residue not frequently found at this position.
The role of threonine at position P1 in the BPTI/Kunitz protease inhibitor family has not yet been well defined (70,71). However, this amino acid residue, which bears a ␤-branched side chain, has been described to bind to the S1 pocket of trypsin in a particularly weak mode (72). In contrast, studies involving the use of peptide combinatorial libraries have suggested that threonine could promote the inhibition of elastases (73). It therefore seems reasonable to suggest that the SmCI-D1 (domain 1, containing such threonine) could contribute to a lesser extent to inhibiting trypsin and be responsible for elas-tase inhibition. The additional presence of a hydrophobic residue at P2Ј, able to interact with the S2Ј subsite of the enzyme, could favor elastase inhibition (74) by this first SmCI domain. The inhibition of pancreatic elastase by the rSmCI D1-D2 and its lack by the rSmCI D2-D3 indicates that the D1 domain is probably responsible for the inhibition of this enzyme, also suggesting an important role of the P1 threonine residue in this event. The important role of P1 threonine residue for elastase inhibition has been demonstrated by the BPTI mutants and is in agreement with our finding. The BPTI K15T mutant (with threonine at P1 position) showed a K i value 3 orders lower than natural BPTI (72).
A remarkable characteristic of the second SmCI domain is the presence of valine at position P1Ј. Indeed, almost all Kunitz domains feature an alanine or glycine residue at this position (69,70), although other small side chain amino acid residues such as aspartate and serine have often been found at this position (67). In these reports it was shown that the presence of ␤-branched side chain residues such as valine causes a decrease in the energy of interaction with trypsin (also described for chymotrypsin), destabilizing the complex (72). Hence, the presence of lysine and glycine at P1 and P1Ј, respectively, of the third domain indicates a likely greater contribution to trypsin inhibition by this domain, a fact confirmed experimentally with the lower K i value determined for rSmCI D2-D3 in comparison with that obtained for rSmCI D1-D2.
BPTI/Kunitz domains, along with those of other inhibitors of the canonical class (Kazal, Bowman-Birk, grasshopper, etc.), can be found as repeats forming a multidomain single chain inhibitor, able to independently interact with several proteases at their reactive sites (68). From this perspective, it is obvious that SmCI can be classified as a multidomain inhibitor of serine proteases with three reactive sites (one at each domain), each one probably being responsible for the preferential inhibition of different serine proteases. It should be emphasized that most inhibitors (including those referred to as polyvalent such as BPTI/Kunitz inhibitors) are specific for proteases belonging to the same mechanistic class (69). Nevertheless, there are also known examples of protein inhibitors that have the capacity to act upon proteases from different mechanistic classes. Generally speaking, if composed of a single domain, these inhibitors will act through nonoverlapping binding sites, although if more than one domain exists, then each domain will be responsible for the inhibition of a given type of protease. For example, ShPI-1, an inhibitor isolated from the sea anemone S. helianthus by the present authors (17), consists of a single domain able to inhibit serine, cysteine, and aspartic proteases. The same applies to an inhibitor of serine and metalloendoproteases isolated from Streptomyces caespitosus (75), an inhibitor of serine and aspartic proteases from Prosopis juliflora (76) and two inhibitors of serine proteases and amylase from barley and wheat (77,78), all of which act through different nonoverlapping binding sites. Bifunctional properties have also been observed in equistatin, an inhibitor with three thyroglobulin I domains, which inhibit cysteine and aspartic proteases (18,19). However, until now the literature lacked the description of a single or multidomain inhibitor such as SmCI with the capacity to inhibit both serine proteases and carboxypeptidases, i.e. SmCI is the first reported inhibitor of metallocarboxypeptidases with BPTI/Kunitz protein structure (a tri-repeated of it, in fact).
The mechanism of most protein inhibitors of CPs, with slight variations (1, 2), depends on the C-terminal amino acid tail of the inhibitory molecule, which docks into the reactive site of the enzyme and loses its last residue during the initial stages of binding, i.e. their behavior resembles an ideal substrate (2,5,7,8,13). Thus, the trimmed residue, Gly 39 in PCI and Glu 66 in LCI (the inhibitors from potatoes and leech, respectively), remains bound to the S1Ј subsite of the enzyme. The carboxylate group of the penultimate residue (Val 38/65 in PCI/LCI) coordinates with the Zn 2ϩ of the active site, affecting S1Ј and S1 subsites, although the two previous residues of the inhibitor (Tyr 37/64 in PCI/LCI, respectively) interact with the S2 subsite (2,5,7,8,13). Thus, the inhibitor affects most CPs sites essential for substrate binding and catalysis. The complexes formed by PCI or LCI with CPs are also strongly stabilized by additional secondary contacts, which in the former case have been evaluated in detail by site-directed mutagenesis (79). Collectively, these studies have indicated the essential role of the C-terminal tail of such inhibitors for enzyme interaction.
In contrast, the C-tail of the SmCI inhibitor here characterized lacks the three to five conserved residues shown by the above mentioned protein inhibitors that interact with CPs in a substrate-like manner (see Fig. 6). The absence of these residues in SmCI, as well as the rigidity and null surface exposure of its C-terminal region (which is involved in a disulfide bridge), suggest that its inhibition mechanism is not linked to this region. This prediction is in agreement with our data indicating that the individual third domain of SmCI, produced by recombinant methods, lacks CPA inhibitory activity but preserves its strong capacity to inhibit trypsin at a level close to that achieved by the whole molecule.
Latexin, a protein inhibitor of CPs found in the nervous system of mammals, is an exception because it interacts with these enzymes in a way that mimics autologous inhibition by the activation segment of pro-carboxypeptidases (12,14). Latexin binds to the top of the enzyme through the interface of its two domains (strands ␤7-␤8), clamping a loop that encompasses residues Asp 273 -Pro 282 of the human CPA4 moiety (14). Remarkably, only a few interactions occur with residues of the enzyme active site (with the S1Ј and S2 subsites), which explains the adaptability of latexin in inhibiting all vertebrate CPs of the A/B-type tested. It is likely that SmCI has a similar mechanism of inhibition to latexin and the propeptide of CPs. The experimental elucidation of the detailed mechanism whereby SmCI acts as a protein inhibitor of CPs, as the first of its kind detected in marine invertebrates, is a pending challenge.
Nevertheless, the minor values of K i against CPA obtained for rSmCI D1-D2 (regarding the tri-domain form) and the lack of this activity in the bi-domain rSmCI D2-D3 and in rSmCI-D3 suggest that the first domain is mainly responsible for CP inhibition, but the whole SmCI structure plays an important role for achieving a higher capacity to inhibit pancreatic-like CPA.
Identifying individual and synergistic elements in the mechanism of inhibition by such domains is of interest in itself and even more in a context of BPTI/Kunitz and metallocarboxypeptidase protein inhibitor types and structures, which had not been found to coexist before. The identification of this new multifunctional inhibitor of proteases reveals the great potential of marine invertebrates as an extensive and still unexplored group of the animal kingdom and for our search for molecules and mechanisms of biological and biotechnological interest. Some of the distinctive properties of marine animals and molecules (such as protease inhibitors in general and the SmCI studied here) are probably related to their moderate level of biological differentiation, their adaptation to the environment, and their feeding and defense mechanisms.